基于标记分水岭算法的高分辨率遥感图像分割方法  

Segmentation of High-resolution Remote Sensing Image Based on Marker-Based Watershed Algorithm

作　　者：孙颖[1] 何国金[1] 

机构地区：[1]中国科学院对地观测与数字地球科学中心

出　　处：《科学技术与工程》2008年第11期2776-2781,共6页Science Technology and Engineering

基　　金：国家自然科学基金(60272032)资助

摘　　要：由于Quickbird等高分辨率遥感图像信息分布的特殊性,一些面向视频(或自然)图像分割方法并不完全适合于高分辨率遥感图像的分割。基于标记的Watershed图像分割算法是一种改进的分水岭算法,它很好地被应用于人脸及其他一些场景图像的分割。把该方法引入高分辨遥感图像的分割,并针对遥感图像的特点,在分割之前采用中值滤波对高分辨率遥感图像进行预处理;同时,在分割过程中采用用小波滤波器替代Butterworth滤波器对梯度图像的低通滤波。不同地物特征的Quickbird图像的分割实验表明,对于纹理比较均一的高分辨率遥感图像,该方法避免了过分割现象,且效率较高;但对于纹理比较复杂的图像,该方法具有一定的局限性。Because of the particularity of high resolution remote sensing image such as Quickbird, the segmentation methods used in common video or image are not completely suitable for high-resolution remote sensing image. The marked-based watershed algorithm is an improvement to the watershed algorithm, and it can be applied well into the segmentation of faces or other scene images. This method was used in the segmentation of high-resolution remote sensing image, and considering the characteristics of remote sensing image, the median filtering was used to carry out to the high-resolution remote sensing image before segmentation; at the in the low-pass filtering of the gradient images in stead of the butterworth filter. same time, wavelet filter was used Segmentation experiments carried out for the Quickbird images with various land covers show that, the method is efficient and it avoids over-segmentation when the texture is homogeneous, however, it also has some limitations when the texture is relatively complex.

关 键 词：图像分割 标记的分水岭算法 高分辨率遥感图像 

分 类 号：P237[天文地球—摄影测量与遥感]